>>>>> In <v9ejs9g05r.fsf@marauder.physik.uni-ulm.de>
>>>>>	Reiner Steib wrote:

>>> in `message-forward-make-body' we have:

[...]

>>> Shouldn't the search for multipart/(signed|encrypted) be restricted to
>>> the headers?

>> I don't think so.  I tested your patch by forwarding again a
>> forwarded message in which the forwarded part contains non-ASCII
>> text and a gpg signature.

> Ah, I didn't think of this case.  Thanks for clarifying.

> Wouldn't any text like "Content-Type: multipart/encrypted" in this
> line give a false match?  Probably this isn't harmful because the
> message body looks nice with `message-forward-make-body-mml', but
> `message-forward-make-body-mime' is safer.  Is this correct?

The complete version is below. :)
The new function `message-signed-or-encrypted-p' dissects the
message to forward and detects a signed or encrypted part by
rightful means.  Even if the function causes an error, Message
chooses the raw mode safely.  Could you test it?  In addition,
I'd like to install it to only the trunk if it is ok.  WDYT?